Business confidence and forecasting of housing prices and rents in

Business confidence and forecasting of housing prices and rents in

Business confidence and forecasting
of housing prices and rents in large German cities
Konstantin A. Kholodilin∗
Boriss Siliverstovs§
November 28, 2013
Abstract
In this paper, we evaluate the forecasting ability of 115 indicators to predict the housing prices and rents
in 71 German cities. Above all, we are interested in whether the local business confidence indicators can
allow substantially improving the forecasts, given the local nature of the real-estate markets. The forecast
accuracy of different predictors is tested in a framework of a quasi out-of-sample forecasting. Its results
are quite heterogeneous. No single indicator appears to dominate all the others for all cities and market
segments. However, there are several predictors that are especially useful, namely the business confidence
at the national level, consumer confidence, and price-to-rent ratios. Even better forecast precision can be
achieved by combining the individual forecasts. On average, the forecast improvements attain about 20%,
measured by reduction in RMSFE, compared to the naı̈ve model. In separate cases, however, the magnitude
of improvement is about 50%.
Keywords: Housing prices; housing rents; forecasting; spatial dependence; German cities; confidence indicators; chambers of commerce and industry.
JEL classification: C21; C23; C53.
∗ DIW
Berlin, Mohrenstraße 58, 10117 Berlin, Germany, e-mail: kkholodilin@diw.de
Zurich,
KOF
Swiss
Economic
Institute,
Weinbergstraße
35,
8092
siliverstovs@kof.ethz.ch
§ ETH
I
Zurich,
Switzerlande-mail:
Contents
1 Introduction
1
2 Data
2
3 Forecasting
5
4 Conclusion
11
References
12
Appendix
14
II
List of Tables
1
List of variables
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2
Business confidence indicators
3
Selection frequency of indicator models into top five forecasting models for each city . . . . . . . 19
4
Rent in primary market: Best forecasting model . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5
Rent in secondary market: Best forecasting model . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6
Price in primary market: Best forecasting model . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
7
Price in secondary market: Best forecasting model . . . . . . . . . . . . . . . . . . . . . . . . . . 23
8
Quarterly year-on-year growth rates in percent: Actual values (2010:1-2013:2) and forecasts for
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2014:2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
9
Forecast accuracy for the training period (2009:1-2013:2) . . . . . . . . . . . . . . . . . . . . . . . 24
List of Figures
1
Secondary market price in large German cities (euros per m2 ), 2004:q1-2013:q3 . . . . . . . . . . 25
2
Secondary market rent for existing housing in large German cities (euros per m2 ), 2004:q1-2013:q3 26
3
Publication schedule of housing prices/rents, DIHK and Ifo business confidence indices . . . . . . 27
4
National and regional business climate indices for construction: Ifo vs. DIHK, 2001:m1-2013:m9
5
Business climate indices of individual cities for construction, 2001:Q1-2013:Q3 . . . . . . . . . . . 29
6
Distribution of the best-forecast indicators, % . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
III
28
1
Introduction
The role of the housing market in the everyday life of society is difficult to overestimate. Housing rents and
prices directly affect the standard of living of every person. In Germany, the housing wealth (6.3 trillions euros
at the end of 2012) accounts for more than a half of wealth of private households (about 12.3 trillions euros).
It is well known that speculative price bubbles on real-estate markets are likely to trigger financial crises,
which can, in turn spill, over to the real economy by producing deep recessions accompanied by huge employment
reductions.
Since the end of 2010, after more than a decade of falling real housing prices, strong rent and especially
price increases have been observed in Germany. This raised doubts and fears in German society. On the one
hand, it is feared that Germany can follow the path of Spain, Ireland, and other bubble countries that ended
in a severe economic crisis1 . On the other hand, the tenants that constitute a majority of German population
are afraid of substantial rent increases that will erode their welfare. The tenants’ discontent takes a form of
massive protests and manifestations endangering political stability in the country. For this reason of the major
issues debated in during recent elections and ongoing coalition negotiations among two leading German parties
CDU/CSU and SPD is the housing policy. Therefore, it is very important to be able to predict the dynamics
of home rents and prices in the nearest future.
There exist very few studies on forecasting housing prices in Germany. an de Meulen et al. (2011) forecast
German real estate prices for four different market segments (new and existing houses and apartments) using
ARDL and VAR as well as forecast combination approaches. Their study is based on monthly data provided by
Immobilienscout24 dating back to 2007. The benchmark is a simple AR model. The authors find that ARDL
and VAR forecasts single-handedly can hardly improve upon the accuracy of AR forecasts, but find some
substantial improvements when weighing the forecasts with the forecast errors of previous periods, especially
for the existing houses segment. The VARs include 26 potential predictors of the real-estate market, namely:
consumer confidence indicators of the European Commission, business confidence indicators of the ifo institute,
and macroeconomic indicators (consumer prices, stock exchange index, mortgage interest rate as well as new
orders and building permits in construction).
1 Deutsche
Bundesbank (2013) in its recent study stated that in several cities the house prices might be overvalued by 5-10%.
1
Kholodilin and Mense (2012) use a panel-data model with spatial effects to forecast the monthly growth rates
of the prices and rents for flats in 26 largest German cities. A big shortcoming of their approach is that their
forecasts are based only on the past growth in the city and in the neighboring city and ignore other indicators
that could contain useful informations about the future price and rent dynamics.
In this paper, we intend to fill this gap and to use alternative predictors in forecasting the housing prices
and rents. In particular, we examine the forecasting performance of the macroeconomic variables, consumer
confidence as well as business confidence indicators. The latter variables, unlike all other, are available not only
at the national but also at the regional level. The regional business confidence indicators are produced by the
local chambers of commerce and industry (CCI) for the whole economy of a region and for its separate branches,
such as industry, construction, services, etc. Despite their potential usefulness, these indicators are neglected
in the literature. To the best of our knowledge, the only paper that takes advantage of the CCI indices for
forecasting purposes is that of Wenzel (2013), who uses business confidence indicators to forecast the economic
growth of German Bundesländer. Such a failure to apply these data can be, among other things, explained by
the difficulty of obtaining them.
The paper has the following structure. Section 2 describes the data used in the paper. Section 3 introduces
the forecasting models and compares their out-of-sample forecast accuracy. Finally, section 4 concludes.
2
Data
This study forecasts four real-estate variables: prices and rents for the existing (secondary market) and newly
built (primary market) housing. The data were provided by a Berlin based research institute empirica 2 that
computes the quarterly housing price/rent indices from the 1st quarter 2004 on. Our data set includes prices and
rents in 71 large German cities from 2004q1 through 2013q2. Thus, the dimension of our sample are N = 71 and
T = 39. Figures 1 and 2 show the dynamics of the housing prices and rents at secondary market, respectively.
Due to a high correlation between the primary and secondary market variables and in order to save space the
graphs of prices and rents at primary markets are suppressed.
The set of potential predictors comprises both macroeconomic variables (15 variables) and confidence indices
2 http://www.empirica-institut.de/empi2007/startseite.html
2
(100 variables). The macroeconomic variables include the housing lending rates and volumes at different loan
maturities as well as the German stock exchange price and performance indices DAX and CDAX, see Table 1.
All of them are available at the national level only and hence are identical for all cities. The macroeconomic
time series were downloaded from the webpage of the Deutsche Bundesbank3 .
The sentiment indices are available both at the national level (business confidence indices of Ifo and consumer
confidence indices of the European Commission) and at the regional level (business confidence indices for
East Germany, Bundesländer or cities). Table 2 lists the national and regional business confidence indices.
“Frequency” refers to the number of times the indicators are published a year. It varies from 2 (semiannual) to
12 (monthly). The vast majority of the CCIs produce their indices at triannual frequency. In some cases, the
surveying and publication frequency has been increased, say, from semiannual to triannual (2-3), or reduced, say,
from quarterly to triannual (4-3). The all-German chamber of commerce and industry (Deutsche Industrie- und
Handelskammertag e.V., or shortly DIHK) collects the data from individual regions and constructs aggregated
indicators for the whole country and four large regions (North, South, East, and West). In addition, Dresden
branch of the Ifo institute conducts it own surveys for East Germany and Saxony. Moreover, the NRW.Bank does
the same for the Bundesland Nordrhein-Westphalen. Furthermore, the sentiment indices of several regions from
the same Bundesland are often aggregated at the Bundesland level (e.g., Niedersachsen and Rheinland-Pfalz).
Figure 3 depicts the publication schedule of housing prices/rents and business confidence indicators. t
corresponds to the 1st quarter of the year, while t − 1 stands for the last quarter of the previous year. It can be
seen that the data on prices/rents are published several weeks later after the end of the reference quarter. The
Ifo indices are typically published on 25th-26th of the reference month. Each quarter sees three Ifo publications:
Ifot,1 is the first month of quarter t, Ifot,2 is the second month of quarter t, and Ifot,3 is the third month of
quarter t. The same publication cycle is valid for the Dresden subsidiary of Ifo and NRW.Bank. Thus, before
the reference quarter ends and much earlier than the price/rent data will be published, some information on the
state of the economy, which may be relevant for predicting the price/rent dynamics, is already available. By
contrast, the DIHK publishes its indices only three times a year: in the beginning of the year (Jahresbeginn),
in the early Summer (Frühsommer), and in the Fall (Herbst). Notice that no data are published in the second
quarter. The exception to this rule are the CCI of Northern Germany (Hamburg, Bremen, and Niedersachsen)
3 http://www.bundesbank.de/Navigation/DE/Statistiken/statistiken.html?nsc=true
3
that publish their business sentiment indices quarterly, and Saarland that produces its indices at monthly
frequency.
Given that the dependent variable has quarterly frequency, while predictors have in many cases a lower
observational frequency, we interpolated such regressors to the monthly frequency by using a linear spline. The
interpolated time series are then sampled at the quarterly frequency, such that March corresponds to the 1st
quarter, June to the 2nd quarter, September to the 3rd quarter, and December to the 4th quarter.
In order to get the time series of the business confidence we contacted all the relevant chambers of commerce
and industry. Unfortunately, we were unable to obtain the sentiment indicators for all the cities of interest.
In some cases, the local CCIs did not respond to our data requests, in other cases, they promised but never
sent the data (like the CCI Nürnberg für Mittelfranken). Therefore, we are very grateful to those CCIs that
supplied us with their indicators. Sometimes we managed to recover the time series from the archives of the
past publications of the business survey indicators. When the business confidence indicators for a city itself
are not available, we are using those of a larger region, to which the city belongs. The latter indicator can
sometimes be even better than the former one. It is known from the anecdotical evidence that in large cities,
such as Berlin and Hamburg, the local construction firms due to their higher costs cannot compete with the
firms coming from the neighbor regions. Hence, the local firms may display lower or even declining business
confidence, in spite of the booming building activity. Thus, using the indices that are based on the opinions of
the local firms can sometimes be misleading.
The business confidence indices used here typically represent the differences between the percentage share
of the positive answers (e.g., the economic situation is good or is going to improve) and the that of the negative
answers (e.g., the economic situation is bad or is going to deteriorate):
Bit = 100 ×
−
A+
it − Ait
−
0
A+
it + Ait + Ait
(1)
−
where A+
it is the number of positive answers given by the firms in the region i in the period t, Ait is the number
of negative answers, and A0it is the number of neutral answers. The index varies between −100 (all firms believe
that the situation is bad) and 100 (all firms believe that the situation is good).
In this study, we utilize four business sentiment indices for forecasting purposes: current situation, future
4
situation (next 12 months), investment plans, and employment plans. When possible those are reported for
the whole economy and for construction industry in particular. Thus, for each region we could have at most 8
different local business confidence indices.
The indices of the current and the future economic situation can be employed to construct a so-called
business climate index:
q
BCI =
current + 100)(B f uture + 100)
(Bit
it
(2)
f uture
current
is the future economic situation index. By
where Bit
is the current economic situation index and Bit
construction, the BCI can take values between 0 indicating extremely bad business climate and 200 pointing to
the excellent business climate.
For some cities only the business climate index is available. Therefore, we computed it also for those cities,
for which we have its components. The BCI is used in the forecasts along with 8 other business confidence
indices.
3
Forecasting
In this section, we describe details of how forecasts of real-estate price indices were made. The four-quarterahead forecasts of the quarterly year-on-year growth rates of the real-estate variables were obtained using a
direct forecasting approach (Marcellino et al., 2006). The forecasts are based on three different specifications
of the forecasting model with gradually increasing information set. Observe that for each city we allow only
one auxiliary indicator to enter the forecasting regression at a time. The first specification contains a single
indicator as the only explanatory variable:
(j)
yt
(j)
where yt
(j)
(j) (j)
(j)
= µi + βi xi,t−4 + it ,
(3)
denotes the quarterly year-on-year growth rate of one of the four real-estate price indices in question
(j)
that is specific to a city (j). The auxiliary indicators are denoted by xi,t , where the super-script (j) allows for
a possibility that some of the indicators are specific to a particular city. Naturally, for national indicators this
5
super-script can be suppressed.
(j)
The second specification of the forecasting model adds own lag of the dependent variable yt−4 as an additional
explanatory variable:
(j)
yt
(j)
(j) (j)
(j) (j)
(j)
= µi + αi yt−4 + βi xi,t−4 + it .
(4)
The third specification of the forecasting model adds a distance-weighted spatial lag of the dependent variable
(W )
yt−4 accounts for spatial correlation between price indices:
(j)
yt
(j)
(j) (j)
(j) (j)
(j) (W )
(j)
= µi + αi yt−4 + βi xi,t−4 + γi yt−4 + it .
(W )
The spatial lag of the dependent variable yt
(5)
was calculated using a spatial weights matrix W such that:
(W )
yt
=
N
X
(j)
wij yt
j=1
A typical element of W is defined as:
Iij d−2
ij
wij = PN
−2
I
k=1 ik dik
(6)
where Iij is the indicator function such that:
Iij =



 1, if dij ≤ d0.25


 0, otherwise
where dij the distance between city i and city j and d0.25 is the first quartile of pairwise distances between all
71 cities.
We elicit the informational content of the auxiliary indicators for the future development of the real-estate
price indices by comparing out-of-sample forecast accuracy of the forecasts models in Equations (3)—(5) with
that of the benchmark models. Correspondingly, for those indicators that are informative about future price
dynamics we should observe substantial increase in forecast accuracy compared to the forecasting performance
of the benchmark models void of this additional information. To this end, we use two benchmark models.
The first benchmark model is a so-called random walk model that uses a historical mean of observed growth
6
rate of the real-estate price indices as a forecast. This model is nested within each of the three specifications
of the forecasting model as it imposes zero restrictions on the slope coefficients in Equations (3)—(5), i.e.,
(j)
αi
(j)
= βi
(j)
= γi
= 0 for all i and j, whenever appropriate. The second benchmark model allows for the lagged
dependent variable to enter the regression. This benchmark model is nested within the models in Equations (4)
(j)
and (5) with the restrictions βi
(j)
= 0 and βi
(j)
= γi
= 0 for all i and j, respectively. Observe that the model
specification in Equation (3) does not nest the autoregressive benchmark model.
The (non-)nested structure of the forecasting and benchmark models has implications on the choice of the
statistical tests for comparing predictive ability of the competing models. In the case of non-nested models
we use the Diebold-Mariano test with the small sample correction proposed in Harvey et al. (1997). When
comparing forecasting accuracy of the nested models we use the test of Clark and West (2007). In both cases
we pairwise tested the null hypothesis of equal predictive accuracy of an indicator-augmented and benchmark
models against an one-sided alternative that the former model produces more accurate forecasts than the latter
model.
In addition, we investigated forecasting performance of various forecast combination schemes (Timmermann,
2006). These include a simple average of all available forecasts (Mean), forecast combinations using weights from
in-sample model fit measured by the Bayesian Information Criterion (BIC), and forecast combinations using
weights derived from the recursively calculated measures of the past forecast performance. In the last group of
forecast combinations the weights are derived from inverse of recursively computed discounted mean squared
forecast errors (MSFE(δ)), where δ denotes a value of chosen discount factor δ = {1, 0.75, 0.50, 0.25} (Watson
and Stock, 2004). We also derived forecast weights by taking average of remaining forecasts after trimming a
certain number of models with the worst forecasting performance (TRIM(τ )), where τ = {0.75, 0.50, 0.25, 0.10}
denotes a quantile in distribution of model-specific MSFEs used as a threshold for discarding models with
the MSFE surpassing this threshold. Last but not least we considered forecast combination based on ranks,
i.e. the forecast weights were computed inversely proportional to model ranking based on the past forecasting
performance in terms of MSFE.
An important aspect of computing forecast combinations, derived from the past forecasting performance,
is that we calculated combination weights based on the information set available at the forecast origin, that
7
is allowing for an appropriate information lag of the target variable when the out-of-sample forecast accuracy
of the models can be evaluated. That is, we simulated information flow to a forecaster under pseudo-real
time conditions. As a result of this setup, forecast combination weights are time-varying as these were recalculated every quarter. For the first few iterations, when the out-of-sample information on forecast accuracy
was not available, we used the equal weighting scheme. This recursive approach to computation of forecast
weights allows us to compare forecasting accuracy of combinations using real-time information versus that of
combinations using full-sample information. In the latter case, forecast combination weights are computed using
the full-sample information and kept constant across all past forecast origins.
We are interested in forecasting dynamics of real-estate price indices four quarters ahead. At the moment of
writing, these prices indices end in 2013:3, implying that we will produce forecasts of the quarterly year-on-year
growth rate for 2014:3. In order to do so, we proceeded in two steps. In the first step, we used a training period
in order to select the best city-specific forecasting model for the chosen forecast horizon. In the second step, we
utilized the identified top-ranked model for producing forecast for 2014:3 for each city.
The training period is from 2009:1 until 2013:3. For each quarter in this period we computed four-quarter
ahead forecasts by appropriately truncating the data set. Due to the fact that after the transformation of the
price indices into the year-on-year growth rates the earliest available observation is for 2005:1, which leaves us
with a rather small estimation sample in order to initialize our forecasting procedure, we used an expanding
estimation window allowing us to use all available observations for estimation of regression coefficients. For
example, the forecast for 2009:1 was produced using estimated coefficients of the model in Equation (??) as well
as two benchmark models for the sample from 2006:1 until 2008:14 . The next forecast for 2009:2 was produced
using estimation results for the 2006:1 until 2008:2, etc.
The results of out-of-sample forecasting using the training period are reported in Table 3 and in Figure 6.
Given a rather large number of alternative models, which makes their pairwise comparison a formidable task,
we summarize the predictive ability of various indicators and their combinations by counting the number of
cities, for which a given indicator was selected among the top five models with the largest forecast accuracy. In
such a comparison we have to distinguish between national and regional indicators. Naturally, all the national
4 Four additional observations were lost due to incorporating the fourth-order lag of the dependent variables in the forecasting
equation.
8
indicators are pertinent to each city in our sample, whereas regional indicators are only relevant for cities in
this particular region. Observe that, even if we make a correction for the smaller number of cities, for which
regional indicators are available, the tentative conclusion is that these regional indicators are of a relatively
minor importance compared to national indicators. The regional indicators are selected most often only once
or twice in the group of top five best indicators. Among the regional indicators, the indicator Region BauGL,
reflecting business climate in the regional construction industry, scores the best, especially in predicting rent in
the primary and secondary housing market segment.
It is interesting to observe that between two benchmark models, the random walk model is selected into the
top-five group much more often than the autoregressive model, reflecting a rather weak informational content
of distant own lags of the growth rates of the price indices at this forecast horizon. Having said this, we observe
that the variables P2R Neubau and P2R Bestand, reflecting the ratio of prices to rents in the primary and
secondary housing segments, appropriately lagged, have the highest selection frequency than any other indicator
for predicting future price dynamics (purchasing prices at primary and secondary markets) four periods ahead.
This finding implies that for predicting future price dynamics the current discrepancy between prices and rents
is more informative than the current growth rates of purchasing prices alone.
Among other indicators, business confidence in construction BUIL.Q2.F6S (Q2: Main factors currently
limiting your building activity, F6S: Other factors) of the European Commission, the index of economic situation
in construction of Ifo, Ifo BauGL and of the Association of German Chambers of Commerce and Industry,
DIHK BauGL, have a relatively high selection frequency. A similar performance is also recorded for selected
indicators from the consumer survey of the European Commission. The highest selection frequency is recorded
for the following indicators based on the corresponding questions CONS.Q1 (Financial situation over last 12
months), CONS.Q2 (Financial situation over next 12 months), CONS.Q6 (Price trends over next 12 months),
CONS.Q10 (Savings at present), and CONS.Q12 (Statement on financial situation of household). It is interesting
to observe that survey question directly asking about intentions about purchasing or building a home within the
next 12 months (CONS.Q14) and home improvements over the next 12 months (CONS.Q15) are not selected
as often as the other above mentioned surveys.
Last but not least, we observe a relatively meager performance of forecast combinations methods. Among
9
all forecast combinations, the one based on the in-sample Bayesian Information Criterion (BIC) has the highest
selection frequency. This conclusion is at odds with the results of an de Meulen et al. (2011), which state the
opposite, emphasizing the important role of forecast combinations in considerable enhancement of predictive
power.
Figure 6 depicts the conditional frequencies of being the best indicator in terms of RMSFE, given that
indicator belongs to one of six groups: business confidence at regional level, business confidence at national level,
consumer confidence, macroeconomic variables, price-to-rent ratios, and forecast combinations. Computing
conditional frequencies allows accounting for a large variation of the number of indicators belonging to each
group: from 2 indicators in the group “price-to-rent ratios” to 63 indicators in the group “business confidence
at regional level”. In all four market segments, forecast combinations have the highest conditional frequency of
producing the best forecasts. In the housing for sale market, they are immediately followed by the consumer
confidence and price-to-rent indicators. In the housing for rent market, the second place is occupied by the
regional business confidence at the level of Bundesländer.
Tables 4—7 contain summary of forecasting performance of the best models selected for each city. We report
the root mean squared forecast error (RMSFE), ratio of RMSFE to that of the benchmark models (random
walk, RW, and autoregressive model, AR), the p-values of the Clark and West (2007) test of equal predictive
ability of the best model and each of the benchmark models. In column Forecast 2014Q3, forecast values of the
quarterly year-on-year growth rates of the price indices for 2014:3 are reported. In columns Mean and St. dev.,
means and standard deviations of actual values of the rent and price growth rates for the period 2010:1-2013:3
are reported.
The summary of forecast and actual values is provided in Table 8, which draws a quite heterogeneous picture
on future price dynamics. On average, we expect that the growth rates will be positive for each price index.
The reported mean growth rate varies from about 2% to 4%, subject, however, to substantial uncertainty across
individual cities that is reflected in values of the reported standard deviations of the forecasts. The forecasts of
prices are more volatile that those of rents with the reported standard deviations of price forecasts are about
twice as large as the standard deviations of rent forecasts, 2.6 (rents at primary market) and 2.3 (rents at
secondary market) versus 5.3 (prices at primary market) and 5.8 (prices at secondary market). The reported
10
correlation between forecasts and past actual values is about 0.6. This indicates that it is very likely that those
cities, for which we observed high growth rates in prices in the past, will continue the trend and in those cities
with stagnating or slowly growing prices the current situation is likely to persist in the near future.
The summary of the forecast accuracy of the best models in the training period is presented in Table 9.
The relative forecast accuracy is measured by the ratio of model-specific RMSFE to that of the RW model.5
The descriptive statistics is calculated using only those models, for which reported RMSFE was numerically
smaller than the RMSFE of the benchmark RW model. The corresponding number of observations is reported
in the row Obs. In parentheses the number of cities, for which the null hypothesis of equal forecast accuracy
with the benchmark random-walk model, was rejected at the 10% significance level by the test of Clark and
West (2007). The number of cities, for which forecast accuracy of the best forecasting model was better than
that of the benchmark RW model, varies from 53 (reported for rent in the primary market) to 62 (reported for
price in the secondary market). According to the results of the Clark and West (2007) test, the null hypothesis
of equal forecasting accuracy of the best model and the benchmark RW model is rejected in 46 out of 53 and
53 out of 59 cases for rent in the primary and secondary markets, respectively, and in 54 out of 57 and 56 out
of 62 cases for price in the primary and secondary markets, respectively. The average decline in RMSFE over
the random-walk model is about 20%, which is of about the same magnitude for all real-estate indices. The
maximum decrease in RMSFE is about 50%, that is again similar across the real-estate indices. The number
of cities for which the benchmark random-walk model produces most accurate forecasts is reported in the row
Obs. (RW). The number of cities, for which no other model was able to produce more accurate forecasts than
the random-walk model, is the largest for rent in the primary market (18), which is about a quarter of cities in
our sample, and the lowest for the price in the secondary market (9).
4
Conclusion
In this paper, we evaluate the forecasting ability of 115 indicators to predict the housing prices and rents in
71 German cities. Above all, we are interested in whether the local business confidence indicators can allow
5 As reported in Tables 4—7 the forecasting performance of the benchmark autoregressive benchmark model was always inferior
to that of the RW model. For example, the AR model was never selected as the best forecasting model. As a result, the RW model
provides the benchmark that is more difficult to improve upon. This is the reason why we compare forecasting performance of the
indicator-augmented models with the RW model.
11
substantially improving the forecasts, given the local nature of the real-estate markets.
In order to test the forecast accuracy of different predictors a four-quarters-ahead out-of-sample forecasting
exercise is undertaken. Its results are quite heterogeneous. No single indicator appears to dominate all the
others. However, there are several predictors that are especially useful, namely the price-to-rent ratios, business
confidence in construction at the national level, and consumer confidence. On average, the forecast improvements
attain about 20%, measured by reduction in RMSFE, compared to the naı̈ve model. In separate cases, however,
the magnitude of improvement is about 50%.
The present analysis utilizes information from national and regional indicators for short-term predicting
real-estate price dynamics. In the future research, the scope of regional or city-specific indicators needs to
be enlarged by collecting local information on factors influencing demand-supply conditions in the real-estate
market such as in-/out-migration, unemployment level, percentage of empty housing, etc.
References
an de Meulen, P., M. Micheli, and T. Schmidt (2011). Forecasting house prices in Germany. Ruhr Economic Papers 0294, Rheinisch-Westfälisches Institut für Wirtschaftsforschung, Ruhr-Universität Bochum, Universität
Dortmund, Universität Duisburg-Essen.
Clark, T. E. and K. D. West (2007). Approximately normal tests for equal predictive accuracy in nested models.
Journal of Econometrics 138 (1), 291–311.
Deutsche Bundesbank (2013). Monatsbericht Oktober 2013.
Harvey, D., S. Leybourne, and P. Newbold (1997). Testing the equality of prediction mean squared errors.
International Journal of Forecasting 13 (2), 281–291.
Kholodilin, K. A. and A. Mense (2012). Forecasting the prices and rents for flats in large German cities.
Discussion Papers of DIW Berlin 1207, DIW Berlin, German Institute for Economic Research.
Marcellino, M., J. H. Stock, and M. W. Watson (2006). A comparison of direct and iterated multistep AR
methods for forecasting macroeconomic time series. Journal of Econometrics 135 (1-2), 499–526.
12
Timmermann, A. (2006). Forecast Combinations, Volume 1 of Handbook of Economic Forecasting, Chapter 4,
pp. 135–196. Elsevier.
Watson, M. W. and J. H. Stock (2004). Combination forecasts of output growth in a seven-country data set.
Journal of Forecasting 23 (6), 405–430.
Wenzel, L. (2013). Forecasting regional growth in Germany: A panel approach using business survey data.
HWWI Research Papers 133, Hamburg Institute of International Economics (HWWI).
13
Appendix
Table 1: List of variables
Code
Description
Source
Frequency
Rent Neubau
Housing rent at primary market
empirica
4
Rent Bestand
Housing rent at secondary market
empirica
4
Price Neubau
Housing price at primary market
empirica
4
Price Bestand
Housing price at secondary market
empirica
4
P2R Neubau
Price-to-rent ratio at primary market
own calculation
4
P2R Bestand
Price-to-rent ratio at secondary market
own calculation
4
BauGL
Current situation in local construction
local CCIs
2 to 12
BauGE
Future situation in local construction
local CCIs
2 to 12
BauGK
Business climate in local construction
local CCIs
2 to 12
BauBeP
Employment plans in local construction
local CCIs
2 to 12
BauInv
Investment plans in local construction
local CCIs
2 to 12
GL
Current situation in whole local economy
local CCIs
2 to 12
GE
Future situation in whole local economy
local CCIs
2 to 12
GK
Business climate in whole local economy
local CCIs
2 to 12
BeP
Employment plans in whole local economy
local CCIs
2 to 12
Inv
Investment plans in whole local economy
local CCIs
2 to 12
Ifo BauGL
Current situation in German construction
Ifo
12
Ifo BauGE
Future situation in German construction
Ifo
12
Ifo BauGK
Business climate in German construction
Ifo
12
Ifo GL
Current situation in whole German economy
Ifo
12
Ifo GE
Future situation in whole German economy
Ifo
12
Ifo GK
Business climate in whole German economy
Ifo
12
DIHK BauGL
Current situation in German construction
DIHK
3
DIHK BauGE
Future situation in German construction
DIHK
3
DIHK BauGK
Business climate in German construction
DIHK
3
DIHK BauBeP
Employment plans in German construction
DIHK
3
DIHK BauInv
Investment plans in German construction
DIHK
3
DIHK GL
Current situation in whole German economy
DIHK
3
DIHK GE
Future situation in whole German economy
DIHK
3
DIHK GK
Business climate in whole German economy
DIHK
3
DIHK BeP
Employment plans in whole German economy
DIHK
3
DIHK Inv
Investment plans in whole German economy
DIHK
3
Region BauGL
Current situation in big region’s construction
DIHK
3
Region BauGE
Future situation in big region’s construction
DIHK
3
Region BauGK
Business climate in big region’s construction
DIHK
3
Region BauBeP
Emploment plans in big region’s construction
DIHK
3
14
Table 1: List of variables (continued)
Code
Description
Source
Frequency
Region BauInv
Investment plans in big region’s construction
DIHK
3
Region GL
Current situation in whole big region’s economy
DIHK
3
Region GE
Future situation in whole big region’s economy
DIHK
3
Region GK
Business climate in whole big region’s economy
DIHK
3
Region BeP
Emploment plans in whole big region’s economy
DIHK
3
Region Inv
Investment plans in whole big region’s economy
DIHK
3
NRW GL
Current situation in whole North Rhine-Westphalia’s economy
NRW.Bank
12
NRW GE
Future situation in whole North Rhine-Westphalia’s economy
NRW.Bank
12
NRW GK
Business climate in whole North Rhine-Westphalia’s economy
NRW.Bank
12
Sachsen BauGL
Current situation in Saxony’s construction
Ifo Dresden
12
Sachsen BauGE
Future situation in Saxony’s construction
Ifo Dresden
12
Sachsen BauGK
Business climate in Saxony’s construction
Ifo Dresden
12
Sachsen GL
Current situation in whole Saxony’s economy
Ifo Dresden
12
Sachsen GE
Future situation in whole Saxony’s economy
Ifo Dresden
12
Sachsen GK
Business climate in whole Saxony’s economy
Ifo Dresden
13
Ostdeutschland BauGL
Current situation in East German construction
Ifo Dresden
12
Ostdeutschland BauGE
Future situation in East German construction
Ifo Dresden
12
Ostdeutschland BauGK
Business climate in East German construction
Ifo Dresden
12
Ostdeutschland GL
Current situation in whole East German economy
Ifo Dresden
12
Ostdeutschland GE
Future situation in whole East German economy
Ifo Dresden
12
Ostdeutschland GK
Business climate in whole East German economy
Ifo Dresden
12
Niedersachsen BauGL
Current situation in Lower Saxony’s construction
CCI Lüneburg-Wolfsburg
4
Niedersachsen BauGE
Future situation in Lower Saxony’s construction
CCI Lüneburg-Wolfsburg
4
Niedersachsen BauGK
Business climate in Lower Saxony’s construction
CCI Lüneburg-Wolfsburg
4
Niedersachsen BauBeP
Emploment plans in Lower Saxony’s construction
CCI Lüneburg-Wolfsburg
4
Niedersachsen BauInv
Investment plans in Lower Saxony’s construction
CCI Lüneburg-Wolfsburg
4
Niedersachsen GL
Current situation in whole Lower Saxony’s economy
CCI Lüneburg-Wolfsburg
4
Niedersachsen GE
Future situation in whole Lower Saxony’s economy
CCI Lüneburg-Wolfsburg
4
Niedersachsen GK
Business climate in whole Lower Saxony’s economy
CCI Lüneburg-Wolfsburg
4
Niedersachsen BeP
Emploment plans in whole Lower Saxony’s economy
CCI Lüneburg-Wolfsburg
4
Niedersachsen Inv
Investment plans in whole Lower Saxony’s economy
CCI Lüneburg-Wolfsburg
4
RLP BauGL
Current situation in Rhineland-Palatinate’s construction
CCI Koblenz
3
RLP BauGE
Future situation in Rhineland-Palatinate’s construction
CCI Koblenz
3
RLP BauGK
Business climate in Rhineland-Palatinate’s construction
CCI Koblenz
4
RLP BauBeP
Emploment plans in Rhineland-Palatinate’s construction
CCI Koblenz
3
RLP BauInv
Investment plans in Rhineland-Palatinate’s construction
CCI Koblenz
3
RLP GL
Current situation in whole Rhineland-Palatinate’s economy
CCI Koblenz
3
RLP GE
Future situation in whole Rhineland-Palatinate’s economy
CCI Koblenz
3
15
Table 1: List of variables (continued)
Code
Description
Source
Frequency
RLP GK
Business climate in whole Rhineland-Palatinate’s economy
CCI Koblenz
3
RLP BeP
Emploment plans in whole Rhineland-Palatinate’s economy
CCI Koblenz
3
RLP Inv
Investment plans in whole Rhineland-Palatinate’s economy
CCI Koblenz
3
Lend.HH.1year.EIR
Effective interest rates of German banks / New business / Housing loans
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
Deutsche Bundesbank
12
to households with an initial rate fixation, floating rate or up to 1 year
Lend.HH.1year.Vol
New business (volumes) of German banks / Housing loans to households
with an initial rate fixation, floating rate or up to 1 year
Lend.HH.1.5year.EIR
Effective interest rates of German banks / New business / Housing loans
to households with an initial rate fixation of over 1 year and up to 5 years
Lend.HH.1.5year.Vol
New business (volumes) of German banks / Housing loans to households
with an initial rate fixation of over 1 year and up to 5 years
Lend.HH.5.10year.EIR
Effective interest rates of German banks / New business / Housing loans
to households with an initial rate fixation of over 5 years and up to 10
years
Lend.HH.5.10year.Vol
New business (volumes) of German banks / Housing loans to households
with an initial rate fixation of over 5 years and up to 10 years
Lend.HH.over10year.EIR
Effective interest rates of German banks / New business / Housing loans
to households with an initial rate fixation of over 10 years
Lend.HH.over10year.Vol
New business (volumes) of German banks / Housing loans to households
with an initial rate fixation of over 10 years
Lend.HH.EIR
Effective interest rates of German banks / New business / Housing loans
to households
Lend.HH.Vol
New business (volumes) of German banks / Housing loans to households
Deutsche Bundesbank
12
Lend.HH.Cost
Effective interest rates of German banks / New business / Housing loans
Deutsche Bundesbank
12
to households (annual percentage rate of charge, total cost of loan)
DAX price
DAX price index / End 1987 = 1000 / End of month
Deutsche Bundesbank
12
DAX performance
DAX performance index / End 1987 = 1000 / End of month
Deutsche Bundesbank
12
CDAX price
CDAX price index / End 1987 = 100 / End of month
Deutsche Bundesbank
12
CDAX performance
CDAX performance index / End 1987 = 100 / End of month
Deutsche Bundesbank
12
BUIL.COF
Construction confidence indicator (Q3 + Q4) / 2
European Commission
12
BUIL.Q1
Building activity development over the past 3 months
European Commission
12
BUIL.Q2.F1S
Factors limiting activity: None
European Commission
12
BUIL.Q2.F2S
Factors limiting activity: Insufficient demand
European Commission
12
BUIL.Q2.F3S
Factors limiting activity: Weather conditions
European Commission
12
BUIL.Q2.F4S
Factors limiting activity: Shortage of labour force
European Commission
12
BUIL.Q2.F5S
Factors limiting activity: Shortage of material and/or equipment
European Commission
12
BUIL.Q2.F6S
Factors limiting activity: Other factors
European Commission
12
BUIL.Q2.F7S
Factors limiting activity: Financial constraints
European Commission
12
16
Table 1: List of variables (continued)
Code
Description
Source
Frequency
BUIL.Q3
Evolution of your current overall order books
European Commission
12
BUIL.Q4
Employment expectations over the next 3 months
European Commission
12
BUIL.Q5
Prices expectations over the next 3 months
European Commission
12
BUIL.Q6
Operating time ensured by current backlog (in months)
European Commission
12
CONS.COF
Consumer confidence indicator (Q2 + Q4 - Q7 + Q11) / 4
European Commission
12
CONS.Q1
Financial situation over last 12 months
European Commission
12
CONS.Q2
Financial situation over next 12 months
European Commission
12
CONS.Q3
General economic situation over last 12 months
European Commission
12
CONS.Q4
General economic situation over next 12 months
European Commission
12
CONS.Q5
Price trends over last 12 months
European Commission
12
CONS.Q6
Price trends over next 12 months
European Commission
12
CONS.Q7
Unemployment expectations over next 12 months
European Commission
12
CONS.Q8
Major purchases at present
European Commission
12
CONS.Q9
Major purchases over next 12 months
European Commission
12
CONS.Q10
Savings at present
European Commission
12
CONS.Q11
Savings over next 12 months
European Commission
12
CONS.Q12
Statement on financial situation of household
European Commission
12
CONS.Q13
Intention to buy a car within the next 12 months
European Commission
12
CONS.Q14
Purchase or build a home within the next 12 months
European Commission
12
CONS.Q15
Home improvements over the next 12 months
European Commission
12
17
Table 2: Business confidence indicators
City
Region
Augsburg
IHK-Bezirk Bayerisch-Schwaben
Berlin
Berlin
Ostwestfalen
Bielefeld
Bochum
Mittleres Ruhrgebiet
Bonn
Bonn/Rhein-Sieg
Bottrop
Nord Westfalen
Braunschweig (only industry)
Braunschweig
Bremen
HK Bremen
Bremerhaven
Bremerhaven
IHK Südwestsachsen
Chemnitz
Cottbus
Südbrandenburg
Dortmund
Ruhrgebiet
Kammerbezirk Dresden
Dresden
Duisburg
Ruhrgebiet
Düsseldorf
Düsseldorf und Mittlerer Niederrhein
Region Nord- und Mittelthüringen
Erfurt
Erlangen
Mittelfranken
Ruhrgebiet
Essen
Frankfurt (all) and IHK-Bezirk Frankfurt (construction)
Frankfurt am Main
Fürth
Mittelfranken
Gelsenkirchen
IHK Nord Westfalen
IHK Bezirk Halle-Dessau
Halle (Saale)
Hamburg
Hamburg
Hannover
IHK-Bezirk Hannover
Heilbronn
IHK Bezirk Heilbronn-Franken
Region Nord- und Mittelthüringen
Jena
Karlsruhe
TechnologieRegion Karlsruhe
Nordhessen (only business climate)
Kassel
Schleswig-Holstein
Kiel
Koblenz
Bezirk der IHK Koblenz
Köln
Stadt Köln
Kammerbezirk Leipzig
Leipzig
Ludwigshafen
Pfalz
Schleswig-Holstein
Lübeck
Magdeburg
Sachsen-Anhalt
Rheinhessen
Mainz
München
Region München
Münster
Nord Westfalen
Mittelfranken
Nürnberg
Oldenburg
Oldenburger Land
Osnabrück - Emsland - Grafschaft Bentheim
Osnabrück
Nordschwarzwald
Pforzheim
Potsdam
Westbrandenburg
Regensburg
Region Oberpfalz-Kelheim
Rostock
IHK-Bezirk Rostock
Saarbrücken
Saarland
Trier
Region Trier
Ulm
IHK-Region Ulm
Wiesbaden
Rhein-Main-Gebiet
Wolfsburg
Lüneburg-Wolfsburg (all), Niedersachsen (construction)
Wuppertal
IHK-Bezirk Wuppertal-Solingen-Remscheid
Würzburg
Mainfranken
Germany and big regions (North, South, West, East)
Saxony and East Germany
North Rhine-Westphalia
Niedersachsen
Rheinland-Pfalz
18
Frequency
3
3
2
2
3
2
4
4
2-4
2-3
3
2
2-3
2
2-3
3
3
2
3
3
2
4
4
4
4
3
4-3
3
4
4-3
3
2-3
3
4
4
3
3
2
3
4
4
2
2
3
3
12
3
4-3
3
4
2-3
3
3
12
12
4
4-3
Source
IHK Schwaben
IHK Berlin
IHK Ostwestfalen zu Bielefeld
IHK Mittleres Ruhrgebiet
IHK Bonn/Rhein-Sieg
IHK Nord Westfalen
IHK Braunschweig
HK Bremen
IHK Bremerhaven
IHK Chemnitz
IHK zu Cottbus
IHK zu Essen
IHK Dresden
IHK zu Essen
IHK zu Düsseldorf
IHK Erfurt
IHK Nürnberg für Mittelfranken
IHK zu Essen
IHK Frankfurt am Main
IHK Nürnberg für Mittelfranken
IHK Nord Westfalen
IHK Halle-Dessau
IHK Hamburg
IHK Hannover
IHK Heilbronn-Franken
IHK Erfurt
IHK Karlsruhe
IHK Kassel-Marburg
IHK zu Kiel
IHK Koblenz
IHK Köln
IHK Leipzig
IHK für die Pfalz in Ludwigshafen am Rhein
IHK zu Kiel
IHK Magdeburg
IHK für Rheinhessen
IHK München und Oberbayern
IHK Nord Westfalen
IHK Nürnberg für Mittelfranken
Oldenburgische IHK
IHK Osnabrück - Emsland - Grafschaft Bentheim
IHK Nordschwarzwald
IHK Potsdam
IHK Regensburg für Oberpfalz/Kelheim
IHK zu Rostock
IHK des Saarlandes
IHK Trier
IHK Ulm
IHK Wiesbaden
IHK Lüneburg-Wolfsburg
IHK Wuppertal-Solingen-Remscheid
IHK Würzburg-Schweinfurt
DIHK
Ifo Dresden
NRW.Bank
IHK Lüneburg-Wolfsburg
IHK Koblenz
19
RN%
2
1
3
11
0
6
3
4
1
2
0
7
0
2
5
3
3
4
1
2
0
9
2
1
2
1
0
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
1
1
1
0
1
0
0
RB%
2
1
4
10
0
9
3
6
2
1
0
11
2
1
5
4
1
3
5
1
2
9
2
0
4
2
0
0
0
0
1
0
0
0
0
0
0
1
0
0
1
0
0
0
2
0
1
0
0
PB%
0
1
1
15
0
2
0
0
1
1
0
7
0
0
0
1
0
0
1
1
0
6
0
1
0
1
0
0
0
1
0
1
0
0
0
0
0
1
0
0
0
0
1
1
0
0
1
0
0
PN%
2
1
1
9
0
2
0
3
1
0
1
6
0
0
0
1
0
5
1
2
1
5
0
1
0
1
0
0
0
1
0
1
0
1
0
0
0
0
0
0
0
0
1
0
0
0
1
0
0
Obs.
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
3
3
3
3
3
3
9
9
9
9
9
9
9
9
9
9
25
25
25
3
3
3
3
Indicator
RLP GK
RLP BauGL
RLP BauGE
RLP BauInv
RLP BauBeP
RLP BauGK
BauGL
BauGE
BauInv
BauBeP
BauGK
GL
GE
Inv
BeP
GK
Ostdeutschland GL
Ostdeutschland GE
Ostdeutschland GK
Ostdeutschland BauGL
Ostdeutschland BauGE
Ostdeutschland BauGK
P2R Neubau
P2R Bestand
Lend.HH.1year.EIR
Lend.HH.1year.Vol
D1Lend.HH.1year.Vol
D4Lend.HH.1year.Vol
Lend.HH.1.5year.EIR
Lend.HH.1.5year.Vol
D1Lend.HH.1.5year.Vol
D4Lend.HH.1.5year.Vol
Lend.HH.5.10year.EIR
Lend.HH.5.10year.Vol
D1Lend.HH.5.10year.Vol
D4Lend.HH.5.10year.Vol
Lend.HH.over10year.EIR
Lend.HH.over10year.Vol
D1Lend.HH.over10year.Vol
D4Lend.HH.over10year.Vol
Lend.HH.EIR
Lend.HH.Cost
Lend.HH.Vol
D1Lend.HH.Vol
D4Lend.HH.Vol
DAX price
D1DAX price
D4DAX price
DAX performance
RN%
0
0
1
0
0
0
2
3
1
0
0
0
2
2
0
2
0
0
0
0
0
0
1
9
8
7
1
6
5
7
0
5
1
0
1
1
1
1
1
7
3
3
1
1
1
4
1
3
3
RB%
0
0
0
0
0
0
3
0
1
1
0
0
2
2
0
2
0
0
0
0
0
0
2
8
5
2
0
5
4
5
0
4
2
2
1
1
1
1
3
12
5
5
0
1
0
3
2
6
2
PB%
0
0
0
0
0
0
1
0
0
0
0
0
1
0
1
0
0
0
0
1
0
0
19
25
5
3
0
5
8
1
1
4
7
3
1
3
3
1
1
2
7
7
5
2
4
4
2
6
2
PN%
0
0
0
0
0
0
0
1
1
0
0
1
1
0
0
0
0
0
0
2
1
0
27
31
4
6
1
3
7
0
1
2
6
2
1
2
4
1
2
5
5
3
4
2
1
4
2
4
1
Obs.
3
3
3
3
3
3
26
26
17
18
27
39
39
23
25
42
11
11
11
11
11
11
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
Indicator
D1DAX performance
D4DAX performance
CDAX price
D1CDAX price
D4CDAX price
CDAX performance
D1CDAX performance
D4CDAX performance
BUIL.COF
BUIL.Q1
BUIL.Q2.F1S
BUIL.Q2.F2S
BUIL.Q2.F3S
BUIL.Q2.F4S
BUIL.Q2.F5S
BUIL.Q2.F6S
BUIL.Q2.F7S
BUIL.Q3
BUIL.Q4
BUIL.Q5
BUIL.Q6
CONS.COF
CONS.Q1
CONS.Q2
CONS.Q3
CONS.Q4
CONS.Q5
CONS.Q6
CONS.Q7
CONS.Q8
CONS.Q9
CONS.Q10
CONS.Q11
CONS.Q12
CONS.Q13
CONS.Q14
CONS.Q15
Mean
BIC
MSFE(1)
MSFE(0.75)
MSFE(0.50)
MSFE(0.25)
TRIM(75)
TRIM(50)
TRIM(25)
TRIM(10)
RW
AR
RN%
1
2
3
1
2
3
2
2
6
1
3
4
4
2
1
5
6
2
2
2
3
2
7
12
5
3
4
11
2
6
3
4
4
7
5
3
2
2
1
1
1
1
2
1
0
0
0
32
1
RB%
3
5
4
1
5
1
3
5
3
3
4
5
4
2
2
4
6
2
3
4
0
2
7
7
6
4
2
8
2
5
3
4
3
2
4
1
4
3
0
1
1
1
1
1
0
0
0
24
1
PB%
3
5
4
1
4
2
1
2
3
0
3
4
5
5
0
18
0
5
1
1
5
0
12
13
0
0
1
2
1
5
1
12
3
15
2
3
1
0
3
0
0
0
0
0
0
1
1
26
1
PN%
0
2
5
1
4
1
0
3
4
0
1
5
3
2
3
10
0
6
1
1
5
1
15
15
1
0
1
6
1
4
3
7
7
10
5
4
1
0
4
0
0
0
0
0
0
0
0
26
1
The entries are selection frequency of each indicator model into top five forecasting models. The column Obs. reports the number of cities for which a given indicator is available.
Indicator
Ifo GL
Ifo GE
Ifo GK
Ifo BauGL
Ifo BauGE
Ifo BauGK
DIHK GL
DIHK GE
DIHK Inv
DIHK BeP
DIHK GK
DIHK BauGL
DIHK BauGE
DIHK BauInv
DIHK BauBeP
DIHK BauGK
Region GL
Region GE
Region Inv
Region BeP
Region GK
Region BauGL
Region BauGE
Region BauInv
Region BauBeP
Region BauGK
Sachsen GL
Sachsen GE
Sachsen GK
Sachsen BauGL
Sachsen BauGE
Sachsen BauGK
Niedersachsen GL
Niedersachsen GE
Niedersachsen Inv
Niedersachsen BeP
Niedersachsen GK
Niedersachsen BauGL
Niedersachsen BauGE
Niedersachsen BauInv
Niedersachsen BauBeP
Niedersachsen BauGK
NRW GL
NRW GE
NRW GK
RLP GL
RLP GE
RLP Inv
RLP BeP
Table 3: Selection frequency of indicator models into top five forecasting models for each city
Obs.
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
71
Table 4: Rent in primary market: Best forecasting model
N
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
City
Aachen
Augsburg
Berlin
Bielefeld
Bochum
Bonn
Bottrop
Braunschweig
Bremen
Bremerhaven
Chemnitz
Cottbus
Darmstadt
Dortmund
Dresden
Duesseldorf
Duisburg
Erfurt
Erlangen
Essen
Frankfurt
Freiburg
Fuerth
Gelsenkirchen
Hagen
Halle
Hamburg
Hamm
Hannover
Heidelberg
Heilbronn
Herne
Ingolstadt
Jena
Karlsruhe
Kassel
Kiel
Koblenz
Koeln
Krefeld
Leipzig
Leverkusen
Ludwigshafen
Luebeck
Magdeburg
Mainz
Mannheim
Moenchengladbach
Muelheim
Muenchen
Muenster
Nuernberg
Oberhausen
Offenbach
Oldenburg
Osnabrueck
Pforzheim
Potsdam
Regensburg
Remscheid
Rostock
Saarbruecken
Salzgitter
Solingen
Stuttgart
Trier
Ulm
Wiesbaden
Wolfsburg
Wuerzburg
Wuppertal
Predictor
D4DAX price
CONS.Q1
D4Lend.HH.1.5year.Vol
CDAX price
D4Lend.HH.1.5year.Vol
CONS.Q10
CONS.Q6
P2R Bestand
D4Lend.HH.1.5year.Vol
Lend.HH.1year.Vol
RW
CONS.Q3
RW
RW
CONS.Q12
RW
RW
Lend.HH.1year.EIR
CONS.Q9
D4Lend.HH.over10year.Vol
BauGL
D1CDAX price
BUIL.Q2.F2S
RW
BUIL.Q2.F7S
CONS.Q11
RW
D4Lend.HH.1year.Vol
Region GE
CONS.Q14
CONS.Q1
CONS.Q6
D4Lend.HH.1.5year.Vol
CONS.Q6
CONS.Q1
BIC
Region BauBeP
RW
Inv
DIHK BauGL
Ifo BauGK
RW
BauGE
RW
CONS.Q3
P2R Bestand
Ifo BauGL
D4Lend.HH.over10year.Vol
RW
DIHK BauGL
D4Lend.HH.1year.Vol
RW
BUIL.Q2.F4S
Ifo BauGL
RW
Lend.HH.1.5year.EIR
CONS.Q2
RW
CONS.Q8
RW
CONS.Q6
CONS.Q10
RW
Region Inv
Ifo BauGL
DAX performance
Lend.HH.1.5year.Vol
RW
RW
BUIL.Q1
CONS.Q11
RMSFE
3.88
1.63
6.55
2.59
2.52
4.64
1.56
2.95
4.46
5.50
2.25
2.17
3.28
2.49
3.35
4.85
3.01
5.16
1.92
2.14
2.29
4.86
2.13
3.27
1.80
1.93
3.29
4.57
4.71
1.46
2.34
3.17
4.21
3.51
2.46
3.88
2.88
1.95
3.93
2.10
1.79
3.11
1.93
2.83
3.66
2.59
1.46
2.02
3.44
1.85
3.27
2.36
1.60
1.66
5.55
2.29
1.22
3.82
0.94
1.88
4.66
2.05
5.13
3.33
1.80
1.88
8.73
4.02
8.21
4.31
0.91
RM SF E
RM SF ERW
0.95
0.52
0.93
0.73
0.87
0.94
0.62
0.91
0.87
0.97
1.00
0.74
1.00
1.00
0.78
1.00
1.00
0.90
0.84
0.83
0.76
0.81
0.86
1.00
0.83
0.84
1.00
0.94
0.90
0.82
0.71
0.71
0.94
0.89
0.77
0.83
0.69
1.00
0.79
0.68
0.84
1.00
0.89
1.00
0.84
0.72
0.86
0.92
1.00
0.64
0.98
1.00
0.76
0.80
1.00
0.72
0.65
1.00
0.78
1.00
0.91
0.77
1.00
0.79
0.62
0.87
0.66
1.00
1.00
0.93
0.81
20
CW
p-value
0.12
0.00
0.00
0.04
0.04
0.17
0.03
0.06
0.13
0.15
.NaN
0.02
.NaN
.NaN
0.01
.NaN
.NaN
0.00
0.02
0.00
0.04
0.02
0.00
.NaN
0.00
0.01
.NaN
0.07
0.12
0.01
0.00
0.02
0.00
0.08
0.03
0.03
0.00
.NaN
0.02
0.02
0.07
.NaN
0.07
.NaN
0.06
0.05
0.08
0.00
.NaN
0.01
0.19
.NaN
0.03
0.02
.NaN
0.00
0.00
.NaN
0.00
.NaN
0.11
0.02
.NaN
0.02
0.04
0.00
0.04
.NaN
.NaN
0.09
0.00
RM SF E
RM SF EAR
0.71
0.42
0.80
0.72
0.79
0.91
0.60
0.81
0.81
0.72
0.96
0.68
0.79
0.86
0.81
0.58
0.86
0.79
0.92
0.81
0.65
0.90
0.82
0.69
0.89
0.80
0.56
0.87
0.75
0.96
0.73
0.62
0.89
0.82
0.67
0.89
0.55
0.78
0.64
0.61
0.82
0.83
0.76
0.87
0.76
0.73
0.80
0.88
0.92
0.49
0.94
0.55
0.75
0.79
0.73
0.69
0.84
0.87
0.66
0.92
0.76
0.69
0.73
0.80
0.70
0.93
0.73
0.84
0.90
0.81
0.86
CW
Forecast
Actual, 2010Q1-2013Q2
p-value
0.02
0.00
0.00
0.05
0.01
0.09
0.02
0.00
0.07
0.03
.NaN
0.00
.NaN
.NaN
0.01
.NaN
.NaN
0.00
0.00
0.01
0.00
0.03
0.00
.NaN
0.02
0.00
.NaN
0.00
0.00
0.04
0.00
0.01
0.04
0.04
0.01
0.12
0.00
.NaN
0.01
0.00
0.08
.NaN
0.02
.NaN
0.01
0.04
0.03
0.00
.NaN
0.00
0.09
.NaN
0.03
0.03
.NaN
0.00
0.00
.NaN
0.00
.NaN
0.03
0.00
.NaN
0.00
0.01
0.02
0.07
.NaN
.NaN
0.11
0.02
2014Q2
0.57
3.58
6.41
0.08
1.11
12.42
0.86
-2.53
2.79
4.09
0.77
1.46
1.74
1.62
4.72
2.55
0.16
8.98
4.07
0.67
4.27
3.82
3.64
-0.07
-0.27
2.09
3.77
0.29
3.72
-0.15
5.74
1.04
3.38
3.40
5.67
6.38
-0.75
0.58
1.42
3.47
0.16
1.11
1.30
2.50
1.12
9.27
2.84
1.18
1.73
4.85
2.09
2.46
2.09
3.41
3.24
6.18
3.45
3.10
2.27
1.12
3.20
5.10
1.30
-0.98
6.03
5.64
9.18
2.26
0.75
3.88
0.04
Mean
3.32
3.79
8.76
3.33
1.70
4.47
1.85
3.80
4.08
6.46
1.69
1.90
2.76
2.52
4.93
3.81
-0.69
7.84
4.65
0.65
3.92
1.50
4.10
-1.27
0.59
2.65
3.98
0.83
3.39
1.38
4.62
2.92
4.52
4.79
4.06
6.12
1.20
1.38
3.05
2.74
1.18
2.54
2.04
3.78
1.79
3.53
1.87
1.80
3.39
4.59
3.10
3.00
1.13
2.77
4.97
3.75
3.10
5.11
2.44
1.79
4.77
2.94
0.93
1.58
3.84
3.17
2.75
4.20
2.28
3.06
0.76
St. dev.
2.52
2.38
3.94
1.84
2.56
4.06
2.34
2.56
3.64
3.88
1.91
2.90
2.73
2.15
0.77
3.50
2.92
2.65
1.70
2.52
2.54
4.67
1.47
2.81
2.12
1.77
2.81
4.85
4.93
1.74
1.80
4.11
4.49
2.39
2.54
2.18
3.68
1.45
4.27
1.80
2.06
1.99
1.60
1.62
4.21
3.38
1.16
1.85
2.73
2.11
3.01
1.58
1.98
1.56
4.75
2.67
0.78
2.58
0.61
1.53
3.87
1.60
4.01
4.02
2.05
2.15
10.63
3.05
7.20
3.49
1.04
Table 5: Rent in secondary market: Best forecasting model
N
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
City
Aachen
Augsburg
Berlin
Bielefeld
Bochum
Bonn
Bottrop
Braunschweig
Bremen
Bremerhaven
Chemnitz
Cottbus
Darmstadt
Dortmund
Dresden
Duesseldorf
Duisburg
Erfurt
Erlangen
Essen
Frankfurt
Freiburg
Fuerth
Gelsenkirchen
Hagen
Halle
Hamburg
Hamm
Hannover
Heidelberg
Heilbronn
Herne
Ingolstadt
Jena
Karlsruhe
Kassel
Kiel
Koblenz
Koeln
Krefeld
Leipzig
Leverkusen
Ludwigshafen
Luebeck
Magdeburg
Mainz
Mannheim
Moenchengladbach
Muelheim
Muenchen
Muenster
Nuernberg
Oberhausen
Offenbach
Oldenburg
Osnabrueck
Pforzheim
Potsdam
Regensburg
Remscheid
Rostock
Saarbruecken
Salzgitter
Solingen
Stuttgart
Trier
Ulm
Wiesbaden
Wolfsburg
Wuerzburg
Wuppertal
Predictor
D4DAX price
CONS.Q2
D4Lend.HH.1.5year.Vol
CDAX price
D4Lend.HH.1.5year.Vol
CDAX price
CONS.Q3
P2R Bestand
D4Lend.HH.1.5year.Vol
CONS.Q8
BUIL.Q2.F6S
CONS.Q3
BUIL.Q2.F7S
D4Lend.HH.1year.Vol
CONS.Q12
RW
CONS.Q2
Lend.HH.1.5year.EIR
CONS.Q15
D4Lend.HH.over10year.Vol
BauGL
D1CDAX price
Ifo BauGL
RW
BUIL.Q2.F7S
Region BauBeP
RW
Region BauGE
CONS.Q4
P2R Bestand
Ifo BauGL
CONS.Q6
D4Lend.HH.1.5year.Vol
CONS.Q6
BUIL.Q2.F2S
P2R Bestand
RW
RW
Inv
RW
Ifo BauGL
D4DAX price
RW
D4DAX price
CONS.Q1
P2R Bestand
Ifo BauGL
D4Lend.HH.over10year.Vol
D4CDAX price
DIHK BauBeP
CONS.Q6
RW
BUIL.Q2.F3S
Ifo BauGL
RW
Lend.HH.1year.EIR
Ifo BauGK
RW
CONS.Q8
CONS.Q13
CONS.Q6
CONS.Q10
RW
CONS.Q7
Ifo BauGL
RLP GE
CONS.Q6
CONS.Q5
RW
DAX performance
CONS.Q6
RMSFE
3.39
1.77
4.09
2.07
2.08
4.34
1.05
2.93
4.91
1.68
1.36
1.42
2.01
3.35
2.42
4.79
2.11
2.33
2.46
2.19
2.41
5.56
1.80
3.08
1.11
1.72
3.54
1.71
3.23
1.80
1.89
1.85
3.98
2.91
1.72
2.47
3.29
2.03
4.07
2.41
1.64
2.41
2.24
1.59
1.32
2.55
1.52
1.94
2.13
2.37
2.96
2.76
1.25
1.70
4.12
1.92
1.17
2.63
1.27
1.57
2.73
1.57
5.55
3.14
1.68
2.17
8.64
3.81
10.17
5.53
1.06
RM SF E
RM SF ERW
0.86
0.57
0.72
0.67
0.81
0.83
0.52
0.88
0.83
0.84
0.98
0.60
0.87
0.96
0.63
1.00
0.96
0.77
0.82
0.89
0.76
0.83
0.75
1.00
0.64
0.93
1.00
0.67
0.93
0.80
0.73
0.57
0.93
0.83
0.79
0.70
1.00
1.00
0.72
1.00
0.81
0.89
1.00
0.81
0.97
0.66
0.75
0.86
0.86
0.75
0.99
1.00
0.77
0.89
1.00
0.70
0.77
1.00
0.82
0.84
0.72
0.67
1.00
0.68
0.60
0.81
0.75
0.99
1.00
0.99
0.85
21
CW
p-value
0.07
0.00
0.02
0.03
0.02
0.01
0.00
0.04
0.11
0.01
0.06
0.00
0.00
0.00
0.01
.NaN
0.18
0.00
0.08
0.00
0.04
0.03
0.01
.NaN
0.00
0.02
.NaN
0.00
0.12
0.00
0.04
0.02
0.00
0.02
0.01
0.00
.NaN
.NaN
0.00
.NaN
0.07
0.01
.NaN
0.00
0.08
0.04
0.06
0.00
0.00
0.00
0.24
.NaN
0.00
0.04
.NaN
0.01
0.08
.NaN
0.00
0.00
0.03
0.02
.NaN
0.01
0.04
0.00
0.06
0.14
.NaN
0.22
0.01
RM SF E
RM SF EAR
0.67
0.53
0.70
0.63
0.75
0.84
0.50
0.79
0.78
0.89
0.95
0.57
0.63
0.91
0.64
0.64
0.86
0.66
0.76
0.79
0.64
0.90
0.75
0.68
0.91
0.84
0.54
0.54
0.85
0.96
0.69
0.59
0.90
0.84
0.72
0.89
0.78
0.78
0.58
0.80
0.76
0.75
0.90
0.74
0.81
0.68
0.70
0.89
0.76
0.57
0.97
0.55
0.77
0.86
0.79
0.72
0.77
0.88
0.68
0.90
0.78
0.68
0.71
0.74
0.65
0.88
0.85
0.89
0.92
0.87
0.91
CW
Forecast
Actual, 2010Q1-2013Q2
p-value
0.02
0.00
0.00
0.03
0.01
0.00
0.00
0.00
0.06
0.02
0.06
0.00
0.08
0.01
0.01
.NaN
0.06
0.00
0.01
0.01
0.01
0.03
0.02
.NaN
0.01
0.01
.NaN
0.04
0.02
0.00
0.03
0.01
0.03
0.04
0.00
0.02
.NaN
.NaN
0.01
.NaN
0.04
0.02
.NaN
0.00
0.04
0.03
0.04
0.00
0.02
0.02
0.16
.NaN
0.00
0.03
.NaN
0.00
0.06
.NaN
0.00
0.00
0.07
0.01
.NaN
0.00
0.01
0.05
0.04
0.02
.NaN
0.12
0.10
2014Q2
0.29
1.47
6.52
-0.75
0.93
0.57
0.29
-3.75
3.06
1.15
1.14
0.77
-1.65
1.36
4.91
1.76
-0.78
7.72
3.59
0.93
4.42
4.10
4.55
0.09
0.88
1.07
3.82
0.72
1.82
3.19
4.41
-0.05
3.09
2.97
4.64
3.64
1.48
0.80
1.02
0.63
1.16
-0.62
0.97
0.28
3.02
9.66
3.05
0.90
-0.07
3.76
2.47
3.24
-0.88
3.40
3.27
4.43
2.84
2.38
2.95
-0.07
2.46
4.02
1.56
-2.99
5.91
3.36
4.65
1.82
2.29
5.26
0.02
Mean
2.98
3.71
7.31
2.34
1.37
3.34
0.68
3.71
4.30
2.01
1.00
0.57
2.01
1.89
4.13
2.27
0.07
4.70
3.37
0.93
4.37
2.19
4.14
-0.95
0.91
1.90
3.62
1.05
3.28
0.93
3.33
1.42
4.14
4.14
3.48
5.08
3.16
1.90
2.72
2.07
1.96
1.78
1.68
1.59
2.47
3.14
2.12
1.66
1.20
4.21
3.68
4.35
0.51
2.96
5.09
3.40
1.97
3.44
3.12
1.06
2.06
2.27
1.55
0.98
3.67
3.21
1.65
3.74
5.71
4.01
0.69
St. dev.
2.49
2.46
3.15
2.03
2.33
4.98
1.89
2.84
4.23
1.71
1.13
2.32
2.05
3.45
1.21
4.13
2.15
1.22
2.12
2.12
2.54
5.20
1.40
2.73
1.58
1.57
3.02
1.73
3.15
2.16
1.84
3.23
4.36
2.28
1.72
1.49
2.31
1.14
5.12
1.05
1.55
2.03
1.86
1.55
1.20
3.78
1.51
2.00
2.37
2.64
2.34
1.49
1.37
1.31
3.04
2.30
1.39
2.17
0.63
1.78
3.56
1.78
4.25
4.60
2.07
2.48
8.24
3.18
8.09
4.45
1.26
Table 6: Price in primary market: Best forecasting model
N
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
City
Aachen
Augsburg
Berlin
Bielefeld
Bochum
Bonn
Bottrop
Braunschweig
Bremen
Bremerhaven
Chemnitz
Cottbus
Darmstadt
Dortmund
Dresden
Duesseldorf
Duisburg
Erfurt
Erlangen
Essen
Frankfurt
Freiburg
Fuerth
Gelsenkirchen
Hagen
Halle
Hamburg
Hamm
Hannover
Heidelberg
Heilbronn
Herne
Ingolstadt
Jena
Karlsruhe
Kassel
Kiel
Koblenz
Koeln
Krefeld
Leipzig
Leverkusen
Ludwigshafen
Luebeck
Magdeburg
Mainz
Mannheim
Moenchengladbach
Muelheim
Muenchen
Muenster
Nuernberg
Oberhausen
Offenbach
Oldenburg
Osnabrueck
Pforzheim
Potsdam
Regensburg
Remscheid
Rostock
Saarbruecken
Salzgitter
Solingen
Stuttgart
Trier
Ulm
Wiesbaden
Wolfsburg
Wuerzburg
Wuppertal
Predictor
Ifo BauGL
P2R Neubau
CONS.Q8
D4DAX price
P2R Bestand
RW
D1Lend.HH.over10year.Vol
P2R Bestand
BUIL.Q2.F6S
P2R Bestand
P2R Bestand
P2R Neubau
RW
CONS.Q15
P2R Neubau
BUIL.Q2.F6S
CONS.Q9
BauInv
CONS.Q10
RW
RW
RW
RW
Lend.HH.1year.EIR
BUIL.COF
RW
Lend.HH.1year.Vol
Lend.HH.5.10year.EIR
P2R Neubau
RW
CONS.Q10
CONS.Q1
P2R Bestand
D4DAX price
RW
RW
RW
P2R Bestand
CONS.Q1
DIHK Inv
Lend.HH.5.10year.EIR
P2R Bestand
DIHK GE
P2R Neubau
P2R Bestand
CONS.Q1
CONS.Q8
CONS.Q2
P2R Neubau
CONS.Q2
CONS.Q1
Lend.HH.1.5year.EIR
Lend.HH.5.10year.Vol
P2R Neubau
P2R Neubau
Ifo GE
Lend.HH.1year.Vol
RW
Region BauGL
CONS.Q1
P2R Bestand
RW
CDAX price
RW
P2R Bestand
BUIL.Q2.F6S
CONS.Q1
CONS.Q1
P2R Neubau
CONS.Q1
P2R Neubau
RMSFE
3.98
3.86
2.39
2.87
1.08
4.16
2.51
5.47
3.97
6.18
3.02
4.92
7.81
3.77
2.02
3.79
2.58
3.09
7.23
4.62
5.20
10.29
5.83
4.24
2.38
34.05
5.83
3.40
4.78
3.51
3.96
3.32
4.69
4.82
1.91
4.71
7.82
1.97
3.95
2.96
4.35
3.23
4.96
4.24
4.95
5.76
3.19
4.05
5.43
7.74
6.28
4.20
1.80
4.97
4.04
5.35
3.77
5.12
4.89
5.07
5.52
5.35
3.86
3.21
2.65
3.00
5.69
4.88
5.96
3.47
3.76
RM SF E
RM SF ERW
0.83
0.78
0.45
0.99
0.82
1.00
0.79
0.74
0.56
1.00
0.74
0.68
1.00
0.91
0.44
0.70
0.70
0.90
1.00
1.00
1.00
1.00
1.00
0.90
0.88
1.00
0.89
0.69
0.76
1.00
0.74
0.71
0.76
0.87
1.00
1.00
1.00
0.62
0.77
0.71
0.92
0.92
0.89
0.65
0.89
0.79
0.64
0.90
0.89
0.77
0.96
0.84
0.88
0.96
0.67
0.86
0.74
1.00
0.74
0.87
0.67
1.00
0.67
1.00
0.50
0.92
0.81
0.75
0.90
0.66
0.83
22
CW
p-value
0.06
0.03
0.00
0.11
0.01
.NaN
0.03
0.01
0.03
0.16
0.00
0.00
.NaN
0.04
0.00
0.00
0.00
0.04
0.02
.NaN
.NaN
.NaN
.NaN
0.04
0.04
.NaN
0.00
0.01
0.02
.NaN
0.01
0.02
0.04
0.06
.NaN
.NaN
.NaN
0.00
0.04
0.01
0.02
0.02
0.01
0.01
0.03
0.01
0.01
0.02
0.00
0.03
0.21
0.06
0.00
0.07
0.00
0.00
0.00
.NaN
0.01
0.00
0.00
.NaN
0.02
.NaN
0.00
0.00
0.00
0.02
0.01
0.00
0.01
RM SF E
RM SF EAR
0.81
0.79
0.84
0.87
0.81
0.86
0.96
0.72
0.76
0.76
0.62
0.69
0.76
0.80
0.40
0.97
0.63
0.91
0.77
0.63
0.93
0.65
0.77
0.84
0.87
0.89
0.83
0.53
0.81
0.80
0.80
0.63
0.77
0.75
0.83
0.58
0.75
0.60
0.69
0.58
0.81
0.85
0.66
0.57
0.89
0.65
0.75
0.71
0.67
0.74
0.71
0.77
0.89
0.80
0.63
0.85
0.75
0.74
0.58
0.64
0.67
0.83
0.57
0.89
0.75
0.80
0.65
0.71
0.83
0.58
0.77
CW
Forecast
Actual, 2010Q1-2013Q2
p-value
0.03
0.01
0.02
0.07
0.00
.NaN
0.03
0.01
0.04
0.05
0.00
0.00
.NaN
0.01
0.00
0.25
0.00
0.05
0.01
.NaN
.NaN
.NaN
.NaN
0.07
0.06
.NaN
0.04
0.00
0.05
.NaN
0.01
0.00
0.07
0.03
.NaN
.NaN
.NaN
0.00
0.04
0.02
0.06
0.00
0.03
0.00
0.02
0.05
0.05
0.07
0.00
0.02
0.05
0.02
0.01
0.01
0.00
0.00
0.02
.NaN
0.05
0.04
0.02
.NaN
0.01
.NaN
0.07
0.02
0.03
0.02
0.01
0.01
0.00
2014Q2
6.32
3.24
9.87
0.58
1.47
2.46
-1.87
2.80
19.77
-4.17
1.76
-3.11
2.63
1.39
3.86
12.58
2.24
5.81
17.33
0.86
3.27
4.90
3.73
4.51
-3.19
7.67
5.94
11.99
5.56
-0.02
10.52
-4.09
3.89
0.58
1.37
1.09
2.48
0.43
6.79
-0.35
8.79
2.18
0.73
-0.15
-4.43
9.18
6.46
3.55
4.54
11.63
6.37
10.43
-0.69
-0.09
1.68
-2.49
5.27
2.51
8.77
6.15
-12.34
0.90
-2.14
-0.59
10.61
6.11
10.03
7.28
-5.09
6.17
-2.63
Mean
3.29
3.80
7.75
1.91
0.17
5.05
-0.07
7.22
6.83
3.39
-0.54
3.11
5.61
0.61
5.34
6.03
1.56
2.93
8.64
3.31
5.13
9.58
6.40
3.78
-0.58
19.11
7.67
3.86
5.64
-0.05
5.11
-0.30
7.92
4.55
2.01
3.04
5.75
0.56
4.98
0.80
2.72
2.21
1.99
4.66
-1.44
5.87
5.13
2.84
4.48
10.68
5.18
4.86
-0.17
3.63
6.57
2.36
4.65
5.48
5.02
3.80
6.59
3.36
2.70
0.86
5.55
4.49
6.75
5.79
2.50
4.60
1.52
St. dev.
3.88
3.79
1.59
1.63
1.00
1.99
3.29
3.28
3.73
4.78
3.51
5.98
6.50
3.84
2.13
3.09
2.47
3.28
3.94
3.06
4.22
8.26
2.73
2.50
2.58
31.55
2.56
3.70
2.84
3.71
2.66
4.75
2.03
4.24
1.40
3.96
6.67
2.26
3.11
2.99
3.87
2.49
5.00
5.25
4.42
6.20
1.97
3.44
4.01
6.20
4.44
3.48
1.67
4.22
1.75
5.80
2.08
1.83
5.18
4.09
6.26
4.33
3.21
1.85
2.68
2.75
5.13
4.03
5.84
3.74
3.96
Table 7: Price in secondary market: Best forecasting model
N
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
City
Predictor
Aachen
Augsburg
Berlin
Bielefeld
Bochum
Bonn
Bottrop
Braunschweig
Bremen
Bremerhaven
Chemnitz
Cottbus
Darmstadt
Dortmund
Dresden
Duesseldorf
Duisburg
Erfurt
Erlangen
Essen
Frankfurt
Freiburg
Fuerth
Gelsenkirchen
Hagen
Halle
Hamburg
Hamm
Hannover
Heidelberg
Heilbronn
Herne
Ingolstadt
Jena
Karlsruhe
Kassel
Kiel
Koblenz
Koeln
Krefeld
Leipzig
Leverkusen
Ludwigshafen
Luebeck
Magdeburg
Mainz
Mannheim
Moenchengladbach
Muelheim
Muenchen
Muenster
Nuernberg
Oberhausen
Offenbach
Oldenburg
Osnabrueck
Pforzheim
Potsdam
Regensburg
Remscheid
Rostock
Saarbruecken
Salzgitter
Solingen
Stuttgart
Trier
Ulm
Wiesbaden
Wolfsburg
Wuerzburg
Wuppertal
RMSFE
P2R Neubau
P2R Neubau
CONS.Q8
RW
P2R Bestand
RW
D1Lend.HH.over10year.Vol
CONS.Q12
BUIL.Q2.F6S
CONS.Q1
P2R Bestand
CONS.Q12
RW
CONS.Q15
P2R Bestand
D4DAX price
CONS.Q5
D4Lend.HH.1year.Vol
CONS.Q10
P2R Bestand
P2R Bestand
CONS.Q10
RW
CONS.Q1
P2R Neubau
RW
Lend.HH.1year.Vol
RW
RW
P2R Bestand
CONS.Q10
P2R Bestand
CONS.Q10
D4DAX price
P2R Neubau
P2R Neubau
CONS.Q10
GE
CONS.Q1
DIHK Inv
BUIL.Q6
P2R Bestand
P2R Neubau
CONS.Q1
RW
CONS.Q1
DIHK BeP
CONS.Q2
P2R Bestand
CONS.Q2
Lend.HH.Cost
Lend.HH.1.5year.EIR
Lend.HH.1year.EIR
P2R Bestand
P2R Bestand
CONS.Q10
D4Lend.HH.1.5year.Vol
RW
Region BauGL
CONS.Q1
P2R Bestand
Ifo BauGL
DAX price
CONS.Q12
D4DAX price
Lend.HH.5.10year.Vol
CONS.Q1
CONS.Q1
P2R Neubau
CONS.Q1
DIHK BauGL
3.71
4.66
3.07
3.23
1.66
3.99
2.75
6.15
3.75
3.97
3.00
7.61
7.63
2.51
2.51
5.36
2.34
2.45
6.40
2.70
5.66
7.11
7.36
3.10
2.94
8.63
7.10
3.39
10.80
3.79
3.25
3.23
4.87
4.99
2.84
3.75
6.49
2.61
4.30
3.34
4.22
2.34
4.35
4.38
7.45
3.62
5.33
2.73
3.43
8.92
4.53
4.86
1.91
3.35
4.63
4.21
3.44
6.00
5.44
2.74
8.02
5.04
3.90
1.87
4.36
4.27
6.06
5.08
7.18
4.59
2.09
RM SF E
RM SF ERW
0.73
0.67
0.49
1.00
0.87
1.00
0.78
0.85
0.57
0.54
0.69
0.82
1.00
0.86
0.57
0.87
0.72
0.80
0.92
0.80
1.00
0.91
1.00
0.82
0.85
1.00
0.91
1.00
1.00
0.89
0.61
0.88
0.78
0.89
0.97
0.65
0.85
0.61
0.80
0.69
0.90
0.81
0.91
0.70
1.00
0.79
0.94
0.84
0.82
0.80
0.65
0.76
0.80
0.81
0.76
0.77
0.78
1.00
0.65
0.63
0.80
0.97
0.78
0.73
0.75
0.92
0.87
0.76
0.86
0.83
0.74
CW
p-value
0.02
0.03
0.00
.NaN
0.05
.NaN
0.04
0.12
0.04
0.00
0.00
0.05
.NaN
0.03
0.00
0.00
0.01
0.01
0.02
0.01
0.14
0.03
.NaN
0.00
0.00
.NaN
0.02
.NaN
.NaN
0.00
0.01
0.02
0.02
0.05
0.10
0.04
0.01
0.01
0.06
0.01
0.07
0.01
0.20
0.04
.NaN
0.02
0.17
0.01
0.00
0.02
0.01
0.04
0.00
0.00
0.00
0.07
0.01
.NaN
0.02
0.00
0.00
0.13
0.01
0.01
0.00
0.00
0.00
0.04
0.02
0.01
0.03
RM SF E
RM SF EAR
0.72
0.81
0.89
0.86
0.84
0.77
0.97
0.80
0.67
0.49
0.64
0.65
0.85
0.77
0.52
0.91
0.64
0.85
0.71
0.78
0.92
0.88
0.85
0.72
0.80
0.96
0.80
0.63
0.40
0.70
0.70
0.88
0.79
0.79
0.77
0.64
0.83
0.56
0.72
0.58
0.81
0.69
0.92
0.60
0.94
0.71
0.88
0.70
0.66
0.86
0.62
0.77
0.79
0.73
0.66
0.63
0.74
0.73
0.55
0.50
0.95
0.83
0.71
0.67
0.78
0.93
0.64
0.75
0.70
0.76
0.79
CW
Forecast
Actual, 2010Q1-2013Q2
p-value
0.00
0.01
0.00
.NaN
0.05
.NaN
0.04
0.02
0.00
0.00
0.00
0.03
.NaN
0.01
0.00
0.00
0.00
0.05
0.02
0.01
0.14
0.00
.NaN
0.03
0.00
.NaN
0.05
.NaN
.NaN
0.01
0.00
0.02
0.00
0.05
0.04
0.02
0.00
0.03
0.02
0.01
0.06
0.00
0.13
0.02
.NaN
0.03
0.09
0.03
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.02
0.00
.NaN
0.02
0.02
0.15
0.07
0.01
0.00
0.00
0.02
0.02
0.01
0.01
0.01
0.01
2014Q2
2.28
8.45
11.61
0.44
1.67
2.22
-3.23
6.03
16.30
1.54
1.37
4.85
2.85
0.39
3.76
7.73
0.13
3.34
15.36
-2.81
1.23
19.12
4.17
2.33
-0.14
1.89
7.04
0.22
0.66
-5.97
11.43
-3.74
15.06
1.13
5.22
7.99
15.83
-0.28
6.63
-0.91
3.44
1.33
4.85
5.97
1.61
7.00
8.20
0.97
4.82
13.17
14.52
13.67
1.12
2.26
1.75
8.45
6.64
2.58
11.90
2.79
-13.60
5.91
-3.35
0.07
9.47
-1.21
9.53
7.76
-1.47
5.61
-2.28
Mean
4.18
6.75
9.21
2.09
0.26
4.84
-0.28
5.24
5.63
4.27
-1.47
5.16
6.06
-0.19
4.68
6.22
1.01
2.60
7.25
1.61
5.24
9.64
7.44
0.66
0.34
6.62
9.24
1.88
4.37
2.76
4.67
-0.25
7.39
5.34
3.88
4.43
6.49
1.09
5.38
0.40
1.22
1.11
3.40
3.75
3.96
4.84
4.27
1.33
1.27
12.40
6.96
6.87
0.50
3.33
6.07
4.17
3.06
5.42
8.23
1.41
5.16
2.40
0.23
-0.15
5.47
4.91
7.64
5.87
4.36
6.14
-0.53
Table 8: Quarterly year-on-year growth rates in percent: Actual values (2010:12013:2) and forecasts for 2014:2
Mean
St. dev.
Minimum
Maximum
Correlation
Rent
primary market
secondary market
actual
forecast
actual
forecast
3.0
2.8
2.6
2.1
1.7
2.6
1.5
2.3
-1.3
-2.5
-1.0
-3.8
8.8
12.4
7.3
9.7
0.62
0.59
23
Price
primary market
secondary market
actual
forecast
actual
forecast
4.2
3.5
4.0
4.3
3.1
5.3
2.9
5.8
-1.4
-12.3
-1.5
-13.6
19.1
19.8
12.4
19.1
0.51
0.65
St. dev.
3.48
4.37
1.82
1.49
1.63
1.71
3.60
4.95
3.48
3.96
3.55
7.56
5.89
2.57
2.60
4.09
1.72
2.64
5.13
2.69
5.09
4.42
4.96
3.18
2.81
5.43
2.67
2.21
10.33
4.15
2.75
3.65
3.20
3.96
1.72
4.32
5.89
2.75
3.23
3.50
4.42
1.93
3.60
5.09
6.60
3.66
3.98
2.31
3.34
6.41
3.62
4.02
1.21
2.71
2.17
4.21
2.25
3.27
5.13
3.38
9.29
4.33
3.10
1.53
3.52
4.57
4.57
4.39
6.61
3.08
2.71
Table 9: Forecast accuracy for the training period (2009:1-2013:2)
primary market
Rent
secondary market
primary market
Price
secondary market
Mean
St. dev.
Minimum
Maximum
0.81
0.10
0.52
0.98
0.79
0.12
0.52
0.99
0.79
0.13
0.44
1.00
0.79
0.11
0.49
1.00
Obs.
Obs. (RW)
53(46)
18
59(53)
12
57(54)
14
62(56)
9
The entries in columns are descriptive statistics of the relative forecast accuracy of the best
models achieved during the forecast training period from 2009:1-2013:2 at the four-quarter
forecast horizon. The relative forecast accuracy is measured by the ratio of model-specific
RMSFE to that of the random-walk model. The descriptive statistics is calculated using only
those models for which reported RMSFE was numerically smaller than the RMSFE of the
benchmark random-walk model. The corresponding number of observations is reported in
the row Obs.. In parentheses the number of cities for which the null hypothesis of equal forecast accuracy with the benchmark random-walk model was rejected at the 10% significance
level by the test of Clark and West (2007). The number of cities for which the benchmark
random-walk model produces most accurate forecasts is reported in the row Obs. (RW).
24
2004 2006 2008 2010 2012
800
25
1500
1700
1500
1300
1300
1200
1000
1200
Oberhausen
Mönchengladbach
Wuppertal
Duisburg
1000
900
2004 2006 2008 2010 2012
2004 2006 2008 2010 2012
1100
1300
Halle
Hagen
Cottbus
Krefeld
2004 2006 2008 2010 2012
Hamm
Salzgitter
Bremerhaven
Wolfsburg
Dortmund
Osnabrück
Bremen
1100
Mülheim
Hannover
Solingen
Bochum
1100
1150
Bielefeld
Saarbrücken
Bottrop
Essen
Bonn
Aachen
Oldenburg
Heilbronn
2004 2006 2008 2010 2012
2004 2006 2008 2010 2012
600
900
Herne
Chemnitz
Gelsenkirchen
Magdeburg
1900
1700
2000
1800
1150 1250 1350 1450
1350
1050
Ludwigshafen
Kiel
Koblenz
Leverkusen
2004 2006 2008 2010 2012
1000
2004 2006 2008 2010 2012
Mainz
Köln
Darmstadt
Münster
2004 2006 2008 2010 2012
2004 2006 2008 2010 2012
1050
1000
Leipzig
Kassel
Braunschweig
Remscheid
Nürnberg
Rostock
Jena
Offenbach
1400
1600
Karlsruhe
Fürth
Mannheim
Potsdam
2004 2006 2008 2010 2012
1250
2004 2006 2008 2010 2012
1200
2004 2006 2008 2010 2012
2004 2006 2008 2010 2012
1200
Dresden
Lübeck
Erfurt
Pforzheim
1200
1400
2004 2006 2008 2010 2012
Ulm
Wiesbaden
Düsseldorf
Erlangen
1600
2400
1600
1800
2004 2006 2008 2010 2012
1400
Berlin
Würzburg
Trier
Augsburg
1300
1600
1900
2004 2006 2008 2010 2012
1100
Frankfurt
Ingolstadt
Heidelberg
Stuttgart
1800
2100
München
Freiburg
Regensburg
Hamburg
2000
3000
4000
Figure 1: Secondary market price in large German cities (euros per m2 ), 2004:q1-2013:q3
2004 2006 2008 2010 2012
7.5
6.5
5.0
6.0
Essen
Dresden
Osnabrück
Solingen
5.0
4.6
4.8
5.0
5.2
4.8
4.6
2004 2006 2008 2010 2012
Gelsenkirchen
Bremerhaven
Chemnitz
4.2
4.4
2004 2006 2008 2010 2012
Remscheid
Duisburg
Herne
Salzgitter
4.4
2004 2006 2008 2010 2012
4.4
4.6
Hagen
Hamm
Leipzig
Cottbus
Halle
Oberhausen
Magdeburg
Dortmund
2004 2006 2008 2010 2012
5.0
Bielefeld
Mönchengladbach
Bottrop
Wuppertal
5.2
5.2
5.0
2004 2006 2008 2010 2012
4.8
2004 2006 2008 2010 2012
5.2
5.6
5.8
2004 2006 2008 2010 2012
5.4
Saarbrücken
Krefeld
Bochum
Rostock
5.4
5.6
5.8
2004 2006 2008 2010 2012
4.8
5.8
5.0
5.0
5.4
Erfurt
Mülheim
Koblenz
Kassel
2004 2006 2008 2010 2012
5.5
6.2
Wolfsburg
Lübeck
Pforzheim
Braunschweig
5.8
6.5
6.0
2004 2006 2008 2010 2012
5.5
Kiel
Hannover
Leverkusen
Ludwigshafen
2004 2006 2008 2010 2012
Berlin
Oldenburg
Aachen
Bremen
5.5
6.0
5.5
6.0
6.0
6.2
6.6
2004 2006 2008 2010 2012
Trier
Fürth
Heilbronn
Augsburg
6.0
6.5
7.0
Potsdam
Mannheim
Nürnberg
Würzburg
2004 2006 2008 2010 2012
7.0
2004 2006 2008 2010 2012
6.5
7.0
7.5
2004 2006 2008 2010 2012
Erlangen
Regensburg
Karlsruhe
Münster
7.0
8.0
7.5
8
7
Jena
Ulm
Bonn
Offenbach
6.5
7.0
7.5
8.0
2004 2006 2008 2010 2012
Ingolstadt
Mainz
Köln
Düsseldorf
6.5
6.5 7.0 7.5 8.0 8.5
Stuttgart
Heidelberg
Darmstadt
Wiesbaden
8.0
9.5
9.0
8.5
München
Frankfurt
Freiburg
Hamburg
9
10 11 12
Figure 2: Secondary market rent for existing housing in large German cities (euros per m2 ), 2004:q1-2013:q3
2004 2006 2008 2010 2012
26
2004 2006 2008 2010 2012
Figure 3: Publication schedule of housing prices/rents, DIHK and Ifo business confidence indices
Pt−1
Pt
DIHKt
Ifot,1
Ifot,2
Pt+1
Pt+2
DIHKt+3
DIHKt+1
Ifot,3
Ifot+1,1
Ifot+1,2
Ifot+1,3
Ifot+2,1
Ifot+2,2
Ifot+2,3
Ifot+3,1
Ifot+3,2
Ifot+3,3
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
t−1
t
t+1
27
t+2
t+3
120
100
80
40
2002 2004 2006 2008 2010 2012 2014
2002 2004 2006 2008 2010 2012 2014
120
60
80
100
Germany Ifo
Germany DIHK
South DIHK
40
40
Germany Ifo
Germany DIHK
North DIHK
X1$Time
X1[, paste("DE_Bau", sIndex, sep = "")]
80
100
120
X1$Time
60
Germany Ifo
Germany DIHK
West DIHK
60
80
60
40
X1[, paste("DE_Bau", sIndex, sep = "")]
Germany Ifo
Germany DIHK
East DIHK
100
120
Figure 4: National and regional business climate indices for construction: Ifo vs. DIHK, 2001:m1-2013:m9
2002 2004 2006 2008 2010 2012 2014
28
2002 2004 2006 2008 2010 2012 2014
Figure 5: Business climate indices of individual cities for construction, 2001:Q1-2013:Q3
140
●
●
●●
●
●
●●
●
120
●
●●
●
●
●
●
●
●
●
●●
100
●
●
●
●
●
80
●
●
●
●●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
60
●
●
●●
●
●
●
●
40
●
●
●
●●
●
●
●
●
●
●●
●
20
●
●
●
●
2001
2003
2005
2007
29
2009
2011
2013
0.10
0.08
BIC.FS
Rent at primary market
Business_conf_region
Business_conf_nation
Consumer_confidence
Macroeconomic
P2R
Combination
0.10
0.08
0.06
0.00
30
CONS.Q14
BUIL.Q6
DAX_price
D1Lend.HH.over10year.Vol
BUIL.Q2.F4S
CONS.Q6
CONS.Q5
CONS.Q11
Ifo_BauGL
D4Lend.HH.1.5year.Vol
Lend.HH.1.5year.EIR
Lend.HH.over10year.EIR
0.04
Lend.HH.5.10year.EIR
0.05
CONS.Q8
0.06
BUIL.Q2.F3S
GE
D4Lend.HH.Vol
CONS.Q15
Region_GE
BauGL
D4Lend.HH.over10year.Vol
P2R_Neubau
RANK.FS
CONS.Q12
CONS.Q10
0.00
BUIL.Q2.F6S
Price at primary market
P2R_Bestand
0.04
CONS.Q1
Business_conf_region
Business_conf_nation
Consumer_confidence
Macroeconomic
P2R
Combination
TRIM.10..FS
BIC.FS
D4DAX_price
Ifo_BauGL
CONS.Q3
Lend.HH.Vol
D4Lend.HH.1.5year.Vol
Lend.HH.1.5year.EIR
D4CDAX_price
Lend.HH.over10year.EIR
Lend.HH.1year.Vol
BUIL.Q6
Lend.HH.over10year.Vol
CONS.Q9
D4Lend.HH.1year.Vol
BUIL.Q2.F6S
0.06
BUIL.Q2.F3S
0.08
GE
Lend.HH.5.10year.EIR
CDAX_price
Lend.HH.1year.EIR
BUIL.Q2.F7S
CONS.Q14
D4Lend.HH.over10year.Vol
CONS.Q8
BeP
CONS.Q2
CONS.Q10
CONS.Q12
Region_GE
P2R_Bestand
CONS.Q1
RANK.FS
P2R_Neubau
0.10
BIC.FS
TRIM.10..FS
RLP_GE
Bayern_BauBeP
Ifo_BauGL
RANK.FS
GE
P2R_Neubau
D4Lend.HH.1.5year.Vol
CONS.Q6
Inv
CONS.Q8
CONS.Q10
Lend.HH.1.5year.EIR
DIHK_BauGL
D1DAX_price
Ifo_BauGK
GL
Region_GE
BUIL.Q2.F3S
CONS.Q12
BUIL.Q2.F6S
D4Lend.HH.1year.Vol
D4Lend.HH.over10year.Vol
CONS.Q9
Lend.HH.1year.EIR
CDAX_price
CONS.Q1
D4CDAX_price
Lend.HH.Vol
D4DAX_price
BUIL.Q2.F4S
Ifo_GK
CONS.Q7
CONS.Q13
CONS.Q4
Region_BauGE
Lend.HH.1.5year.Vol
0.12
BIC.FS
RLP_GE
TRIM.10..FS
CONS.Q6
NRW_GE
Ifo_BauGL
D4Lend.HH.1.5year.Vol
RANK.FS
D4Lend.HH.over10year.Vol
CONS.Q13
BauGL
Inv
D4Lend.HH.1year.Vol
P2R_Neubau
Lend.HH.1year.EIR
CDAX_price
CONS.Q1
BUIL.Q2.F4S
CONS.Q7
GE
Region_GE
CONS.Q8
P2R_Bestand
CONS.Q9
Lend.HH.over10year.EIR
D4CDAX_price
CONS.Q2
Lend.HH.Vol
CONS.Q3
D4DAX_price
CONS.Q11
Ifo_GK
D1CDAX_price
DIHK_BauGL
Ifo_GE
Figure 6: Distribution of the best-forecast indicators, %
0.07
Price at secondary market
Business_conf_region
Business_conf_nation
Consumer_confidence
Macroeconomic
P2R
Combination
0.03
0.02
0.02
0.01
0.00
0.14
Rent at secondary market
Business_conf_region
Business_conf_nation
Consumer_confidence
Macroeconomic
P2R
Combination
0.06
0.04
0.04
0.02
0.02
0.00